def test_Corpus_interacao(self):
# teste com description.inf
c = Corpus(
"/home/iamjabour/workspacePessoal/tese/corpus/machine_learning_tableextraction_paper2/webtablegrnd/html"
)
print "path", c.path
print "parser", c.parser
print "proof", c.proof
print "documents len", len(c.documents)
x = 0
while True:
d = c.getDocument()
x += 1
if d == None or x > 0:
break
print d.path
def __init__(self, corpusPath, configFile=None, extractors=[], limit=int(2**31-1), output=None):
if not configFile:
if os.path.exists('config_example.cnf'):
config = Configurator('config_example.cnf')
else:
config = Configurator('apps/config_example.cnf')
else:
config = Configurator(configFile)
if output == None:
output = sys.stdout
self.output = output
self.metric = config.metric()
self.marker = self.metric.marker()
self.corpus = Corpus(corpusPath)
self.extractors = extractors
self.limit = limit
self.benchmark = []
def __init__(self, corpusPath, configFile=None, extractors=[], limit=int(2**31-1), output=None, pfilenames=False, initial = 0):
if not configFile:
# verifica se existe o arquivo de configuracao padrao e o carrega
if os.path.exists('config_example.cnf'):
config = Configurator('config_example.cnf')
else:
config = Configurator('apps/config_example.cnf')
else:
# carrega o arquivo de configuracao fornecido
config = Configurator(configFile)
if output == None:
# imprime todas as informacoes na saida padrao
output = sys.stdout
if not limit:
limit = int(2**31-1) # roda em todo o corpus
# somente imprimir o id do arquivo com seu path caso True
self.pfilenames = pfilenames
self.output = output # saida padrao onde sera impresso os benchmark
self.metric = config.metric() # metrica que sera utilizada
self.marker = self.metric.marker() # marcador que sera utilizado
self.corpus = Corpus(corpusPath) # carrega o corpus a partir do path
if self.corpus == None:
return None
self.extractors = extractors # extratores que serao utilziados
self.limit = limit # configura o limite para o que foi escolhido
self.initial = initial
# inicializa uma lista vazia onde serao guardados os resultados
self.benchmark = []
class Benchmark:
def __init__(self, corpusPath, configFile=None, extractors=[], limit=int(2**31-1), output=None, pfilenames=False, initial = 0):
if not configFile:
# verifica se existe o arquivo de configuracao padrao e o carrega
if os.path.exists('config_example.cnf'):
config = Configurator('config_example.cnf')
else:
config = Configurator('apps/config_example.cnf')
else:
# carrega o arquivo de configuracao fornecido
config = Configurator(configFile)
if output == None:
# imprime todas as informacoes na saida padrao
output = sys.stdout
if not limit:
limit = int(2**31-1) # roda em todo o corpus
# somente imprimir o id do arquivo com seu path caso True
self.pfilenames = pfilenames
self.output = output # saida padrao onde sera impresso os benchmark
self.metric = config.metric() # metrica que sera utilizada
self.marker = self.metric.marker() # marcador que sera utilizado
self.corpus = Corpus(corpusPath) # carrega o corpus a partir do path
if self.corpus == None:
return None
self.extractors = extractors # extratores que serao utilziados
self.limit = limit # configura o limite para o que foi escolhido
self.initial = initial
# inicializa uma lista vazia onde serao guardados os resultados
self.benchmark = []
def process(self, debug=False):
"""
Executa o benchmark, amarzenando os resultados parciais em
self.benchmark. para visualizar o resultado de forma adequada utilize
pprint()
"""
count = self.initial
while True:
count += 1
doc = None
if debug:
print count, self.limit
if self.limit != int(2**31-1) and count <= self.limit:
doc = self.corpus.getDocument(id=count)
elif self.limit == int(2**31-1):
doc = self.corpus.getDocument()
else:
break
if not doc or doc == None:
break
try:
if debug:
print doc.id, doc.path
except:
pass
proof = self.corpus.getProof(doc)
result = (0.0,0.0)
for extractor in self.extractors:
if not self.pfilenames:
self.marker.reset()
extractor.process(doc.content, self.marker)
extracted = self.marker.labels
result = self.metric.process(extracted, proof)
self.benchmark.append(result)
def pprint(self):
"""
Consolida o benchmark e imprime o resultado na saida padrão
"""
result = {}
if len(self.benchmark) == 0:
print 'Nao foi gerado nenhum resultado'
return {}
for key in self.benchmark[0]:
result.update({key:[0,0,0]})
for id,doc in enumerate(self.benchmark):
print id, doc
for key in doc:
result[key][0] += doc[key][0]
result[key][1] += doc[key][1]
result[key][2] += doc[key][2]
print
print 'total'
print result
r = {}
for k in result:
if result[k][1] > 0:
if result[k][2] > 0:
precision = result[k][0]/float(result[k][2])
else:
precision = 1
recall = result[k][0]/float(result[k][1])
if (recall+precision) > 0:
f = (2*recall*precision)/(recall+precision)
else:
f = 0.0
r.update({k: [recall, precision, f]})
else:
recall = 1
precision = 1 if result[k][2] == 0 else 0
f = (2*recall*precision)/(recall+precision)
r.update({k: [recall, precision, f]})
print
print >> self.output, 'key\trecall\tprecision'
print >> self.output, r
return r
class Benchfile:
def __init__(self, corpusPath, configFile=None, extractors=[], limit=int(2**31-1), output=None):
if not configFile:
if os.path.exists('config_example.cnf'):
config = Configurator('config_example.cnf')
else:
config = Configurator('apps/config_example.cnf')
else:
config = Configurator(configFile)
if output == None:
output = sys.stdout
self.output = output
self.metric = config.metric()
self.marker = self.metric.marker()
self.corpus = Corpus(corpusPath)
self.extractors = extractors
self.limit = limit
self.benchmark = []
def _process(self, doc):
proof = self.corpus.getProof(doc)
result = (0.0,0.0)
for extractor in self.extractors:
extractor.process(doc.content, self.marker)
extracted = self.marker.labels
result = self.metric.process(extracted, proof)
self.benchmark.append(result)
def process(self, file=False):
"""
Executa o benchmark, amarzenando os resultados parciais em
self.benchmark. para visualizar o resultado de forma adequada utilize
pprint()
"""
if file:
doc = self.corpus.getDocument(file)
print "process Document:", doc.path
print
self._process(doc)
return
count = 0
while True:
count += 1
doc = self.corpus.getDocument()
if doc == None or count > self.limit:
break
self._process(doc)
def pprint(self):
"""
Consolida o benchmark e imprime o resultado na saida padrão
"""
result = {}
for key in self.benchmark[0]:
result.update({key:[0,0,0]})
for id,doc in enumerate(self.benchmark):
print id, doc
for key in doc:
result[key][0] += doc[key][0]
result[key][1] += doc[key][1]
result[key][2] += doc[key][2]
print
print 'total'
print result
r = {}
for k in result:
if result[k][1] > 0:
if result[k][2] > 0:
precision = result[k][0]/float(result[k][2])
else:
precision = 1
recall = result[k][0]/float(result[k][1])
r.update({k: [recall, precision]})
else:
r.update({k: [1, 1 if result[k][2] == 0 else 0]})
print
print >> self.output, 'key\trecall\tprecision'
print >> self.output, r
return r